Insulin aspart has been shown, in medium-term studies, to achieve reductions in HbA(1c) without increasing the risk of major hypoglycaemia compared with pre-meal human insulin. The aim of the present 3-year study was to evaluate the long-term safety and efficacy of insulin aspart in people with type 1 diabetes. This was a 30-month extension of a multinational, multicentre, open-label, parallel-group study of 753 people with type 1 diabetes, originally randomly allocated to treatment with insulin aspart or unmodified human insulin before meals, with NPH insulin as basal insulin. Main outcomes measures were hypoglycaemia (major or minor), adverse events and HbA(1c). As insulin aspart became commercially available in some countries before the end of the trial, analyses of HbA(1c) used 30-month data to maintain statistical power. The relative risk estimate of major hypoglycaemia was similar between treatment groups (relative risk [RR] 1.00 [95% CI 0.72, 1.39]). The risk of having a minor hypoglycaemic episode was higher with insulin aspart than with human soluble insulin (RR 1.24 [1.09, 1.39] p=0.024). Insulin aspart was significantly superior to human insulin with respect to overall glycaemic control, with a baseline-adjusted HbA(1c) difference of -0.16 (-0.32, -0.01)% (p=0.035). Insulin aspart was well tolerated and effective during long-term treatment. The HbA(1c) advantage was maintained with insulin aspart without any adverse impact on the rate of major hypoglycaemia.